[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/sci/ - Science & Math

View post   
File: 26 KB, 709x602, 1530789343904.png [View same] [iqdb] [saucenao] [google]
9905967 No.9905967 [Reply] [Original]

Realistically how far away are we from superintelligent AI?

>> No.9905971
File: 69 KB, 1024x576, 1506205371175.jpg [View same] [iqdb] [saucenao] [google]
9905971

~2045

>> No.9905975

What do you mean? AI is already superhuman on general intelligence.

>> No.9905979
File: 436 KB, 1930x1276, HLAIpredictions.png [View same] [iqdb] [saucenao] [google]
9905979

How accurate is this article?

https://arxiv.org/pdf/1705.08807.pdf

>> No.9905985

>>9905967
We don't even have basic AI. AI is a science fiction mechanic that can't be created by humans in the real world. It is however, a great way to chase grants and high paying jobs. Seems retarded normies have no fucking clue about anything. lol

>t. "AI" programmer

>> No.9906298 [DELETED] 

>>99059853
Neural networks is AI

>> No.9906299

>>9905985
Neural networks is AI

>> No.9906402

>>9906299
There are no actual neural networks in existence that are human-created.

>> No.9906424

>>9905967
It is much closer than the contrarians are willing to admit.

>> No.9906489

>>9906402
The weight and biases are automatically adjusted through back propagation. But that doesn't mean the network isn't human created

>> No.9906896

>>9906299
neural networks are a way to justify using millions of free parameters in an empirical model. they are purely empirical tools.

they are not going to do anything resembling reasoning any time soon.

they can be used as "black boxes" in larger systems developed by humans to carry out computations in more complex reasoning, but this reasoning is still designed by humans.

there is a difference between the field of ai, which is mostly about building empirical models on high dimensional data, and ai from science fiction/sentience etc.

>> No.9906899

>>9906896
although i guess huge nets used in natural language (chatbots etc) processing may develop internal representations which mirror human reasoning. i dont know much about this tho. if anyone knows more please share.

>> No.9906940

>>9905967
AI is a meme and will never happen. If you remember, even in the 90's everybody shilled that AI is right at the horizon. Needless to say, we still don't have it yet

>> No.9906968

>>9906940
Go away thunderf00t.

>> No.9906975

>>9906940
>It doesn't exist, therefore it will never exist

>> No.9906977 [DELETED] 

>>9906896
The human brain is an empirical model with even more free parameters. Yet we can reason just fine

>> No.9906978

>>9906896
The human brain is an empirical model with even more free parameters. Yet we can reason just fine. The brain is also a "black box"

>> No.9907140

>>9906978
While the electric mechanisms in the brain are quite well understood, the results on chemical mechanisms are only broadly known. Furthermore, biological neurons include temporal variables.
There's no evidence that just adding more neurons to current neural network models could give birth to intelligence and >consciousness similar to humans.
I think most AI researchers would agree that current methods are qualitatively insufficient for scifi AI.

>> No.9907147

>>9906968
Has thunderf00t ever said AI is a meme?

>> No.9907243

>>9907140
We have neural networks with temporal variables, they're called recurrent neural networks. AI is artificial intelligence which current neural networks already have. What current neural networks don't have is general intelligence or consciousness. The goal of AI is general intelligence, not necessarily a consciousness agent

>> No.9907250

>>9907140
>>9907243
There's also no way to know if an AI is conscious until we solve the hard problem of consciousness